Neural networks in the banking business: study of the influence of exogenous factors

Abstract

Due to a number of weaknesses of the mathematical models found in use in the banking industry, the author proposes the use of new methods such as the «automatic generation of a trained neural network». The neural network simulates outgoing consolidated banking indicators based on the input of a number of economic and demographic indicators associated with the country of the banks’ location. This new approach was designed to research the influence of input factors in the neural network on a single output factor. It was tested using data of twelve (12) European countries and researching the influence of a few selected exogenous (economic and demographic) indicators on the «Percentage of Bank Non-performing loans». The results may be used as empirical evidence for the eligibility of the proposed method.